NewroMap: Mapping CNNs to NoC-interconnected Self-Contained Data-Flow Accelerators for Edge-AI
NewroMap: Mapping CNNs to NoC-interconnected Self-Contained Data-Flow Accelerators for Edge-AI
复制标题
NewroMap：将 CNN 映射到 NoC 互连的独立数据流加速器，以实现边缘 AI
DOI：
--
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Lennart Bamberg
中科院分区：
文献类型：
--
作者：
J. Joseph;M. S. Baloğlu;Yue Pan;R. Leupers;Lennart Bamberg
传统的AI加速器受到边缘工作负载的冯-诺依曼瓶颈的限制。特定领域的加速器（通常是神经形态的）通过应用近/内存计算，NoC互连的多核设置和数据流计算来解决这个问题。这需要神经网络的有效映射（即网络层到核心的分配）来平衡资源/内存，计算和NoC流量。在这里，我们为主要的卷积神经网络（CNN）引入了一种名为Snake的映射。它通过将层折叠到空间相邻的核心来利用CNN的前馈性质。对于MobileNet和ResNet，我们实现了高达3.8倍的总NoC带宽改进与随机映射。此外，还提出了NewroMap，它通过元启发式继续优化Snake映射;它还模拟NoC流量，并可以与TensorFlow模型一起工作。通信进一步优化，与模拟中显示的纯snake映射相比，延迟改善高达22.52%。
Conventional AI accelerators are limited by von-Neumann bottlenecks for edge workloads. Domain-specific accelerators (often neuromorphic) solve this by applying near/in-memory computing, NoC-interconnected massive-multicore setups, and data-flow computation. This requires an effective mapping of neural networks (i.e, an assignment of network layers to cores) to balance resources/memory, computation, and NoC traffic. Here, we introduce a mapping called Snake for the predominant convolutional neural networks (CNNs). It utilizes the feed-forward nature of CNNs by folding layers to spatially adjacent cores. We achieve a total NoC bandwidth improvement of up to 3.8× for MobileNet and ResNet vs. random mappings. Furthermore, NewroMap is proposed that continues to optimize Snake mapping through a meta-heuristic; it also simulates the NoC traffic and can work with TensorFlow models. The communication is further optimized with up to 22.52% latency improvement vs. pure snake mapping shown in simulations.
Ratatoskr：用于在 3D NoC 中进行深入功耗、性能和面积分析与优化的开源框架
DOI：
10.1145/3472754
发表时间：
2021
期刊：
ACM Trans. Model. Comput. Simul.
影响因子：
--
作者：
Jan Moritz Joseph;Lennart Bamberg;Imad Hajjar;Behnam Razi Perjikolaei;Alberto García-Ortiz;Thilo Pionteck
通讯作者：
Thilo Pionteck